Imctools

Looking for ways to analyze your dataset? Need help with a software package?
Computational image analysis of multiplexed Imaging Mass Cytometry acquisitions
RaulCatena
Posts: 30
Joined: Mon Nov 27, 2017 12:49 pm

Re: Imctools

Post by RaulCatena » Tue Jul 17, 2018 9:35 pm

Santhosh,

Linked to the reply to the other thread, and to point 2. Yes, you can apply the compensation in h++ (or other adjustments) and then I recommend you export the channels one by one (but you can do batches of images to make it fast, for instance, export 3 channels for 100 images would be 3 batch exports) as white JPG. That will work down the line in Ilastik/CP. Additionally, a manual with instructions for segmentation directly in h++ will be released.

Best,
Last edited by RaulCatena on Tue Jul 17, 2018 9:44 pm, edited 1 time in total.

ssivajothi
Posts: 31
Joined: Thu Jan 11, 2018 9:00 pm

Re: Imctools

Post by ssivajothi » Tue Jul 17, 2018 9:41 pm

Hi Raul,

Thank you for the replies. Yes I believe I understand. I will try segmentation with Ilastik/Profiler from H++ exports. Also looking forward to the manual to do it right on H++. Do you expect any difference in segmentation quality via these two methods (using Vito's python scripts vs segmentation on H++)?

Santhosh

RaulCatena
Posts: 30
Joined: Mon Nov 27, 2017 12:49 pm

Re: Imctools

Post by RaulCatena » Tue Jul 17, 2018 9:46 pm

h++ has two procedures. A) A watershed with a few tricks B) Pixel classification in a similar way to what Ilastik does.

A is simpler but inferior, so for optimal results, Vito's approach is best. That being said, when you go down the road of pixel classification, the quality depends on the ground truth provided, so make sure you patiently train good pixels :)

User avatar
votti
Posts: 24
Joined: Mon Nov 27, 2017 12:46 pm

Re: Imctools

Post by votti » Wed Jul 18, 2018 8:36 am

ssivajothi wrote:
Tue Jul 17, 2018 1:42 pm
Hi Vito,

I am a beginner with Python and IMCtools, just a couple of questions:

1) Is IMCtools package supposed to be 'installed' via terminal or Jupyter notebook? Or am I supposed to copy it somewhere and call it? Sorry if terminology I'm using is wrong
2) is there a way to use histocat++ in this pipeline? For example: take txt/mcd file into histocat++, perform compensation, generate single tiffs for each channel and then take this into your segmentation pipeline? Can you clarify if this is possible or the preferred order for doing it? If I do something like this, which middle part of the script do I start from in your pipeline?

Thank you,
Santhosh
Hi Santhosh,

To:
1) imctools can be installed from a terminal via: `pip install git+https://github.com/BodenmillerGroup/imctools.git`
When I next find time I will update the documentation to clarify this.
I would recommend using a environment/package manager like Conda (https://conda.io/docs/) to manage your python and installed python packages.

2) Unfortunately I am not a Mac user and have no experience with HistoCAT++. I hope Rauls comments clarified things

ssivajothi
Posts: 31
Joined: Thu Jan 11, 2018 9:00 pm

Re: Imctools

Post by ssivajothi » Wed Jul 18, 2018 3:21 pm

Hi Vito,

Thank for that, I got it working. Yes, Raul's comments clarified some things about histocat++.

Does "old_example_preprocessing_pipeline.ipynb" contain all steps from start to finish of the pipeline?

Sorry for many beginner questions.

Thanks,
Santhosh

User avatar
votti
Posts: 24
Joined: Mon Nov 27, 2017 12:46 pm

Re: Imctools

Post by votti » Fri Jul 20, 2018 11:03 am

ssivajothi wrote:
Wed Jul 18, 2018 3:21 pm
Hi Vito,

Thank for that, I got it working. Yes, Raul's comments clarified some things about histocat++.

Does "old_example_preprocessing_pipeline.ipynb" contain all steps from start to finish of the pipeline?

Sorry for many beginner questions.

Thanks,
Santhosh
Hi Santhosh,

No problem, I recognize that I should work on some better documentation, thus I am happy for feedback what's currently not clear. I hope I find time to do it sometimes soonish...

As the name suggests, "old_example_preprocessing_pipeline.ipynb" is rather an old version. Thus I would rather recommend to stick to the pipline at: https://github.com/BodenmillerGroup/Imc ... rial.ipynb together with the matching Cellprofiler pipelines: https://github.com/BodenmillerGroup/Imc ... _pipelines
However the other one also seems fairly complete.

Cheers, Vito

Post Reply